On Sat, Aug 1, 2009 at 6:17 PM, Martijn van Oosterhout wrote:
> On Sat, Aug 01, 2009 at 09:37:11AM +0200, PFC wrote:
>> Actually, I think (see the bottom of my last email) that this would be
>> a
>> good argument for the per-column COLLATE patch...
>
> Standard SQL COLLATE support is per col
On Sat, Aug 01, 2009 at 09:37:11AM +0200, PFC wrote:
>> PFC writes:
>>> - for short strings (average 12 bytes), sort is CPU-bound in strcoll()
>>> - for longer strings (average 120 bytes), sort is even more CPU-bound in
>>> strcoll()
>>
>> No news there. If you are limited by the speed of text co
PFC writes:
- for short strings (average 12 bytes), sort is CPU-bound in strcoll()
- for longer strings (average 120 bytes), sort is even more CPU-bound in
strcoll()
No news there. If you are limited by the speed of text comparisons,
consider using C locale.
regards,
PFC writes:
> - for short strings (average 12 bytes), sort is CPU-bound in strcoll()
> - for longer strings (average 120 bytes), sort is even more CPU-bound in
> strcoll()
No news there. If you are limited by the speed of text comparisons,
consider using C locale.
rega
There was a thread some time ago about sorting... it kind of died...
I did some tests on a desktop (Postgres 8.3.7, kubuntu, Core 2 dual core,
4GB RAM, RAID1 of 2 SATA disks)
Quick conclusions :
- grabbing the stuff to sort can be IO bound of course (not here)
- for short strings (average 12
